ARTICLE

DOI: 10.1038/s41467-018-06703-2 OPEN A therapeutic approach to pantothenate kinase associated neurodegeneration

Lalit Kumar Sharma1,2,4, Chitra Subramanian1, Mi-Kyung Yun3, Matthew W. Frank1, Stephen W. White 3, Charles O. Rock 1, Richard E. Lee 2 & Suzanne Jackowski 1

Pantothenate kinase (PANK) is a metabolic enzyme that regulates cellular coenzyme A (CoA) levels. There are three human PANK genes, and inactivating mutations in PANK2 lead

1234567890():,; to pantothenate kinase associated neurodegeneration (PKAN). Here we performed a library screen followed by chemical optimization to produce PZ-2891, an allosteric PANK activator that crosses the blood brain barrier. PZ-2891 occupies the pantothenate pocket and engages the dimer interface to form a PANK•ATP•Mg2+•PZ-2891 complex. The binding of PZ-2891 to one protomer locks the opposite protomer in a catalytically active conformation that is refractory to acetyl-CoA inhibition. Oral administration of PZ-2891 increases CoA levels in mouse liver and brain. A knockout mouse model of brain CoA deficiency exhibited weight loss, severe locomotor impairment and early death. Knockout mice on PZ-2891 therapy gain weight, and have improved locomotor activity and life span establishing pantazines as novel therapeutics for the treatment of PKAN.

1 Department of Infectious Diseases, St. Jude Children’s Research Hospital, Memphis, TN 38105, USA. 2 Department of Chemical Biology and Therapeutics, St. Jude Children’s Research Hospital, Memphis, TN 38105, USA. 3 Department of Structural Biology, St. Jude Children’s Research Hospital, Memphis, TN 38105, USA. 4Present address: Nurix, Inc, 1700 Owens Street, Suite 205, San Francisco, CA 94158, USA. Correspondence and requests for materials should be addressed to S.J. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2

rare, life-threatening neurological disorder known as Here, we report the development of a drug capable of allos- Apantothenate kinase-associated neurodegeneration terically activating the alternate PANK isoforms as a potential (PKAN) arises from mutations in the human PANK2 PKAN therapeutic. The lead pantazine, PZ-2891, arose from the gene1 leading to a prominent extrapyramidal movement disorder LipE-guided chemical optimization of a hit from a high- and a characteristic deposition of iron in the basal ganglia2–4. throughput screen designed to identify inhibitors and activators Pantothenate kinase (PANK, EC 2.7.1.33) is the first and rate- of PANK333,34. Due to the high cooperativity of the PANK controlling step in the only pathway for coenzyme A (CoA) dimer9 the binding of PZ-2891 to one protomer locks the biosynthesis. CoA is a major acyl group carrier in biology and opposite protomer in a constitutively active state that is refractory participates as a key cofactor and regulator of intermediary to feedback inhibition by acyl-CoA. PZ-2891 crosses the blood metabolism5. Three genes express four closely-related mamma- brain barrier to elevate brain CoA. A mouse model of CoA lian isoforms: PANK1α, PANK1β, PANK2, and PANK35. The deficiency employing the neuron-selective deletion of PANK1 isoform expression level and potent feedback inhibition by acyl- and PANK2 genes exhibits weight loss, severely impaired loco- CoAs control the intracellular CoA content6–10. The physiological motor activity, and early death. PZ-2891-treated knockout mice evaluation of PANK knockout mice and mice treated with a gain weight, and have improved locomotor activity and life span. PANK inhibitor establish a key role for the intracellular CoA These data suggest pantazines as a novel approach to treating concentration in supporting oxidative metabolism, ketone utili- PKAN. zation and glucose homeostasis11–14. PKAN pathology is thought to arise from neuronal CoA deficiency. This view is strengthened by the recent discovery that mutations in CoA synthase cause a Results similar neurodegenerative disease15. The PANK2 gene is abun- LipE-guided optimization identifies PZ-2891 as a PANK dant in human neuronal tissues and the majority of the mutations modulator. Prior attempts at hit-to-lead optimization of hits associated with PKAN result in the expression of truncated or from a large high throughput screen for modulators of PANK3 mutant PANK2 proteins with little or no catalytic activity16. failed to generate suitable chemical leads due to flat structure- Human PANK2 is localized to the mitochondrial intermembrane activity relationships and poor solubility34. To address these space17,18. Mouse PanK2 was reported to be mitochondrial19,20, shortcomings, we reevaluated the hit list using the alternative but a mouse PanK2 mitochondrial targeting sequence has not approach of filtering for compounds with both lead-like mole- been identified, and others report a cytosolic localization18,21. cular weight (<350) and lipophilic ligand efficiency (LipE > 2). −/− − Pank2 mice do not recapitulate the neurological PKAN LipE (pIC50 cLogP) blends both potency and lipophilicity to phenotype13,22. quantitatively compare the molecules, and ranks them based on There are no disease-modifying treatments for PKAN4, and physicochemical properties suitable for biophysical therapeutics are desperately needed. One approach is to use characterization35,36. These efforts prioritized a piperazine urea protected phosphopantothenates to bypass PANK and elevate hit, PZ-2789 (Fig. 1a), which was subjected to a hit-to-lead CoA23. Phosphopantetheine, CoA, or S-acetyl- optimization using LipE as the primary driving metric to create a phosphopantetheine have been suggested as PKAN modulators compound series called pantazines (Fig. 1a). PANK activation in based on their ability to reverse hopantenate inhibition of CoA cells (see below) requires a high-affinity pantazine, and a PANK – synthesis24 30. All these approaches use compounds with physi- inhibition assay without acetyl-CoA was used to rank the pan- cochemical properties that suggest they will not penetrate the tazines. The structure–activity relationships revealed the impor- 31,32 blood brain barrier , and protected phosphopantothenates tance of a small branched alkyl group at R1, a carbonyl (H-bond have no effect on mouse brain CoA levels23. acceptor) next to the piperidine ring, and an electron

a

HN N N CN N N CN N N CN O N O N O NN

PZ-2789 (hit from screen) PZ-2863 PZ-2891

IC50 (PANK3) = 844 nM IC50 (PANK3) = 122 nM IC50 (PANK3) = 1.3 nM ClogP = 3.3 ClogP = 3.3 ClogP = 2.3 LipE = 2.8 LipE = 3.6 LipE = 6.6 Solubility (pH 7.4) = 1.4 μM Solubility (pH 7.4) = 25 μM Solubility (pH 7.4) = 23 μM

T1/2 (mice plasma) = 17 min Oral Bioavailibity (F) = 46%

bcPotency: CH2 > O > NH LipE: CH2 > NH > O Potency: Br > Cl > CN Inactive LipE: CN > Cl > Br O

R1 X N N R2 N N CN O N NN (Branched alkyl required) PZ-3067 Additional N-atom Potency: s-Bu > i-Pr > t-Bu Required for improves activity μ LipE: i-Pr > s-Bu > t-Bu activity IC50 (PANK3) > 100 M (negative control)

Fig. 1 Chemical progression of pantazines. a Lipophilic ligand efficiency (LipE = pIC50 – clogP) guided optimization of the pantazine series from the initial hit (PZ-2789) to PZ-2891. b Summary of the structure-activity relationships of >60 compounds synthesized during the optimization of the pantazine scaffold. c Chemical structure of an inactive pantazine, PZ-3067, which is used as a negative control in cellular assays

2 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE

withdrawing group at R2 for the activity of the compounds (Fig. 2g) shows that pantothenate and PZ-2891 both form (Fig. 1b). The LipE-guided optimization culminated in the hydrogen bonds with R207, and the isopropyl substituent of PZ- identification of PZ-2891 that was ∼800 times more potent than 2891 docks into the hydrophobic cavity occupied by the dimethyl the initial hit. The improved potency and reduced lipophilicity group of pantothenate. However, pantothenate does not interact (~1 log unit reduction in cLogP) resulted in a significantly with the dimer interface, and the connection between R306′ and improved LipE for PZ-2891 (Fig. 1a). The three key changes from T209 in the pantothenate complex is mediated by a water the initial hit to PZ-2891 were: replacement of the t-butyl group molecule rather than by a direct hydrogen bond. The structural with isopropyl, switching the urea linker with an acetamide lin- analysis suggests that PZ-2891 first binds to the pantothenate site ker, and substitution of the nicotinonitrile side chain with of the PANK3•ATP•Mg2+ complex followed by closing of the pyridazine-3-carbonitrile. Compared to the screening hit, PZ- flexible loop, and the engagement and re-organization of the 2891 exhibited superior adsorption, distribution, metabolism, and dimer interface. The induced fit mechanism is the most common excretion properties that indicated it could be used for proof of binding mode observed in high-affinity inhibitors37, and with PZ- principle experiments in cells and animals (Fig. 1a; Supplemen- 2891 the two protomers become locked together to create a tary Tables 1, 2). PZ-3067 (Fig. 1c), an inactive isomer of PZ- structure that is distinct from the structures of all the normal 2891, was selected as a biochemical control compound for catalytic intermediates9. mechanism of action studies (Supplementary Table 1).

+ Pantazines are allosteric PANK activators. The idea behind the Pantazines bind to the PANK3•ATP•Mg2 complex. The lead deployment of pantazines as PANK activators in cells was to pantazine, PZ-2891, inhibited PANK3 with nM affinity, whereas exploit the high cooperativity of the enzyme to lock it in an active the inactive PZ-3067 had no effect (Fig. 2a). PZ-2891 inhibited all conformation that cannot be inactivated by acetyl(acyl)-CoA. three human and mouse pantothenate kinase isoforms (Supple- This unusual effect was predicted from the established interac- mentary Fig. 1). We used a less potent pantazine, PZ-2724 (IC 50 tions between PANK, its substrates, acetyl-CoA inhibitor and = 1.1 μM) to determine how pantazines interrupted the ordered pantazine outlined in Fig. 3a. At subsaturating pantazine con- kinetic mechanism of PANK3 (Supplementary Fig. 2). Kinetic centrations, a new drug-induced catalytic cycle exists that pre- analysis of PANK3 with respect to ATP showed that PZ-2724 was vents PANK3 from returning to a conformation that can bind to, an uncompetitive inhibitor resulting in a decreased Vmax and and be inhibited by, acetyl-CoA. Within the cell, the PANK3 ATP Km consistent with the pantazines binding to the PAN- + dimer exists in one of two distinct conformations: the inactive K3•ATP•Mg2 complex. Pantazine inhibition with respect to conformation stabilized by the binding of acetyl-CoA and the pantothenate was noncompetitive. The conclusion that panta- + + active conformation stabilized by the binding of ATP•Mg2 .In zines bind tightly to the PANK3•ATP•Mg2 complex was cor- the normal catalytic cycle, ATP cooperatively binds to the PANK roborated by thermal stabilization studies showing that PZ-2891 dimer switching both protomers to the active, closed conforma- stabilized PANK3 only in the presence of ATP (Fig. 2b). Both tion. Pantothenate binds and catalysis proceeds through the three, ATP and ATP + PZ-2891 stabilization of PANK3 to thermal structurally characterized intermediates, and both of the active denaturation were concentration-dependent and saturable, sites empty9. The phosphopantothenate product is rapidly con- whereas PZ-2891 alone was not (Supplementary Fig. 3). Also, verted to CoA by the biosynthetic pathway. Acetyl-CoA is a [3H]ATP completely dissociated from PANK3 during gel filtra- feedback inhibitor of PANK, and cellular acetyl-CoA levels rise tion chromatography, but when PZ-2891 was added to the + until almost all of the PANK exists in the acetyl-CoA-bound, sample, the PANK3•[3H]ATP•Mg2 •PZ-2891 complex was sta- + inactive protein conformation. bilized and isolated (Fig. 2c). Analysis of the PANK3•ATP•Mg2 Sub-saturating PZ-2891 concentrations interrupt this normal •PZ-2891 complex stability by surface plasmon resonance showed catalytic cycle and feedback regulatory mechanism. In the in vitro a long residence time (34 min), and a calculated affinity of 0.2 nM biochemical assays to optimize the pantazine series, acetyl-CoA (Fig. 2d). was absent and the presence of ATP (>1 mM) ensured that PANK3 existed only in the active conformation. Under these PZ-2891 binds across the PANK dimer interface to stabilize conditions, PZ-2891 only acts as an inhibitor by titrating the the active PANK conformation. The crystal structure of the active sites (Fig. 3a). However, the presence of acetyl-CoA in cells PANK3•AMPPNP•Mg2+•PZ-2891 complex (Supplementary means that upon completion of the catalytic cycle PANK3 is Table 3; Supplementary Fig. 4) reveals that PZ-2891 binds across available to bind acetyl-CoA and switch to the inactive the dimer interface to simultaneously interact with both PANK3 conformation. When PZ-2891 is bound to only one protomer, protomers (Fig. 2e). The ‘ring of ligands’ linking the two proto- the opposite protomer remains capable of catalysis. However, at mers explains the high thermal stability of the complex (Fig. 2e). the end of the pantazine-dependent catalytic cycle, PANK3 The PZ-2891 interactions are described with residues on the remains locked in its active conformation by the pantazine opposite protomer indicated with a prime (Fig. 2f, g). The iso- preventing binding of the acetyl-CoA inhibitor and allowing propyl moiety packs into a hydrophobic cavity created by V250′, another cycle of catalysis to proceed (Fig. 3a). This effect renders I253′, Y254′, Y258′, and A269′ located on the flexible flap that is PANK3 dimers with PZ-2891 bound to only one protomer disordered in the PANK3•AMPPNP•Mg2+ complex9 but refractory to feedback inhibition by acetyl-CoA (Fig. 3a). becomes structured in the PANK3•AMPPNP•Mg2+•PZ-2891 The unusual effect of PZ-2891 acting as both an orthosteric complex (Fig. 2f, g). The carbonyl group forms hydrogen bond inhibitor and an allosteric activator of PANK3 activity in the interactions with R207, which explains the lack of PZ-3067 presence of acetyl-CoA was tested in biochemical assays designed binding (Fig. 1c). The piperazine ring acts as a spacer to present to mimic the mixture of ligands present in cells. The assays the pyridazine ring to the opposite protomer to hydrogen bond contained either PANK3 or PANK3 plus PZ-2891 at a with R306′ and engage W341′ by π-π stacking interactions. PZ- concentration of drug that partially inhibited (25%) the total 2891 interaction with R306′ also allows an inter-protomer PANK3 activity. In the absence of PZ-2891, PANK3 activity was hydrogen bond between R306′ and T209 to further stabilize the extinguished in a concentration-dependent manner by acetyl- dimer. Comparison of the PANK3•AMPPNP•Mg2+•PZ-2891 CoA as is normally observed6 (Fig. 3b). However, in the presence complex to the PANK3•AMPPNP•Mg2+•pantothenate complex of PZ-2891 approximately half of the PANK3 activity that

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2

a e AMPPNP

100 PZ-3067 PZ-2891

50

PZ-2891 % Maximum activity % Maximum

0 0.00 0.05 0.10 Pantazine (μM) PZ-2891 AMPPNP b 20 f PANK3 20 °C +ATP

°C 7 °C

Δ PZ-2891 – ATP 15 R207 PZ-2891 + ATP α ′ T209 9 S195 10 AMPPNP ′ R306 PZ-2891 E138 5 ′ S303′

%Fluorescence/ L338 Δ 0 0 1020304050607080 A269′ Temperature (°C) Y258′ W341′ L263′ ′ β13′ c 30 80 A337 Y254′ + PZ-2891 I253′

) α ′ ′ 4 – PZ-2891 10 V268 V250′

mAU 60 OD 20 α ′

280 7

40 (mAU)

10 20

H]ATP (DPM × 10 H]ATP g 3 [ 0 0 6 8 10 12 14 R207 Elution (ml) α ′ T209 9 S195 AMPPNP d 50 R306′ Water K = 203 ± 0.05 pM D Pan E138 40 L338′ S303′

30

W341′ 20 ′ Water A269 Y258′ A337′ L263′ β13′ Response (RU) 10 Y254′ α10′ I253′ ′ 0 V268 V250′ α ′ 0 100 300 500 7 Time (s)

Fig. 2 PZ-2891 binds to the PANK3•ATP•Mg2+ complex with nM affinity. a PZ-2891 (n = 10) (filled circles) and PZ-3067 (open circles) (n = 2) inhibition of PANK3. The data were fit to the Morrison equation and 1.3 ± 0.2 nM was the consensus PZ-2891 IC50 from five experiments using different batches of purified enzyme. A representative data set is shown. b Thermal stabilization of PANK3 by 8 mM ATP, 8 μM PZ-2891 or 2 mM ATP plus 2 μM PZ-2891. Control (red line), PZ-2891 (black line), ATP (green line), and ATP plus PZ-2891 (blue line). The peaks of the first derivative plots of the thermal denaturation curves identify the temperatures at which 50% of the protein is unfolded. The shifts in the thermal denaturation temperatures were calculated from experiments performed in triplicate, and the averages rounded to the nearest degree. c Isolation of the PANK3•[3H]ATP•Mg2+•PZ-2891 3 complex by gel filtration chromatography. Black trace, PANK3 protein elution profile (A280); Red trace, [ H]ATP elution profile in the presence of PANK3 + PZ2891; Green trace (on baseline), [3H]ATP elution profile with PANK3 alone. One example of duplicate experiments is shown. d Surface plasmon resonance analysis of PZ-2891 binding to PANK3 in the presence of 1 mM ATP•Mg2+. The data (black lines) were fit to a 1:1 binding model (orange lines). 6 -1 -1 –4 -1 The association (ka), dissociation (kd), equilibrium (KD) constants were calculated as 2.37 × 10 M s , 4.82 × 10 s and 0.203 nM respectively. The residence time (1/kd) of PZ-2891 was 34 min. Data are the average of two experiments in duplicate. e Overview of the PANK3 dimer illustrating that PZ- 2891 binds across the dimer interface. The two PANK3 protomers are colored cyan and gold, PZ-2891 is purple and AMPPNP is green. f Close-up view of PZ-2891 bound across the dimer interface of the PANK3•AMPPNP•Mg2+ complex illustrating the key hydrogen bonding interactions (dotted red lines) with both PANK3 protomers. g Structure of the PANK3•AMPPNP•Mg2+•pantothenate (Pan) complex (PDB ID: 5KPR) with the pantothenate hydroxyl rotated into the catalytically active position

4 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE

ab 100

+ PZ-2891 Acetyl-CoA 50 % Maximal activity

0 ADP 050100 Acetyl-CoA (μM)

c 1000 Pan 45 uM Pan 45 uM + PZ-2891 800 Pan 90 uM P-Pan ATP Pan 90 uM + PZ-2891 600

400 (pmoles)

200 Phosphopantothenate Pan 0 PZ-2891 0 10 20 30 40 Time (min) d DMSO CoA +PZ-2891

Pan % Maximum

0 50 100 150 200 Position (mm)

Fig. 3 Activation of PanK by PZ-2891. a Kinetic model illustrating how PZ-2891 binding activates PANK3. PANK3 is a highly-cooperative enzyme where both active sites exist in either the active or inactive conformation. The inactive ‘open’ conformation (red rectangles) is stabilized by the binding of acetyl- CoA, and the active ‘closed’ conformation (green or curved shape) is stabilized by ATP binding. In the normal catalytic cycle, PANK3 returns to a ligand- free state that allows it to bind acetyl-CoA and switch to the inactive conformation. PZ-2891 binds tightly to the active, ATP-bound PANK3 conformation. When a fraction of PANK3 is occupied by PZ-2891, the catalytic cycle is operating on only one protomer of the dimer. In this catalytic cycle, the active monomer of PANK3 empties, but remains locked in the active conformation by ATP•Mg2+•PZ-2891 binding to the other protomer. The drug-induced catalytic cycle prevents PANK3 from adopting the open, inactive conformation making PANK3 refractory to feedback regulation by acetyl-CoA. Pan pantothenate, P-Pan phosphopantothenate. b PZ-2891 renders PANK3 refractory to acetyl-CoA inhibition. PANK3 (1 μg/assay) activity at different concentrations of acetyl-CoA was determined with ATP (1 mM) in the presence (open red circles) and absence (filled circles) of 2.5 μM PZ-2891. c The time course assays contained PANK3 (1 μg/assay), ATP (1 mM) and acetyl-CoA (100 μM) such that PANK3 was inhibited about 95% by acetyl-CoA in the absence of pantazine. The addition of 2.5 μM PZ-2891 to the assay increased the PANK3 rate in a pantothenate-dependent manner. Filled black circles are 45 μM pantothenate, filled blue squares are 90 μM pantothenate, red open circles are 45 μM pantothenate plus 2.5 μM PZ-2891, and green open squares are 90 μM pantothenate plus 2.5 μM PZ-2891. Data in b and c are from two independent experiments that were performed in duplicate. d C3A cells were radiolabeled with [3H]pantothenate in the presence (red trace) or absence (black trace) of 10 μM PZ-2891 for 24 h. The cells were extracted, and the labeled metabolites were analyzed by thin-layer chromatography and imaged with a Bioscan detector. Representative chromatograms from two independent experiments in duplicate are shown remained in the presence of PZ-2891 was refractory to acetyl- pantothenate concentration to twice its Km further accelerated CoA inhibition (Fig. 3b). These data directly support the PANK3 activity (Fig. 3c). These data are consistent with existence of the drug-induced catalytic cycle that prevents the pantazines not affecting the pantothenate Km (Supplementary enzyme from interacting with acetyl-CoA. A second experiment Fig. 2c), and predict that pantothenate concentrations above the to model the cellular environment employed PANK3 whose Km for PANK would provide the highest level of pantazine- activity was repressed >95% by the presence of acetyl-CoA dependent PANK activity in the presence of acetyl-CoA. These (Fig. 3c). Under these conditions, an increase in pantothenate data predict that pantazine-treated cells would have elevated CoA from 45 μMto90μM had little impact on the rate because and reduced pantothenate due to the activation of PANK. C3A PANK3 existed in a conformation unable to bind the substrate. cells were labeled with [3H]pantothenate in the presence and The addition of PZ-2891 activated PANK3, and increasing the absence of 10 μM PZ-2891 and the CoA and pantothenate levels

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 were determined following separation by thin-layer chromato- (Fig. 3c). A reduction in the pantothenate in the cell culture graphy (Fig. 3d). PZ-2891 treatment increased the amount of medium reduced PZ-2891 activation of CoA synthesis, whereas intracellular CoA and lowered intracellular pantothenate, thus increasing the media concentration potentiated the PZ-2891 illustrating the operation of the pantazine-dependent catalytic effect (Fig. 5a). Pantothenate itself had no effect on CoA content, cycle that functions to elevate CoA in cells (Fig. 3a). consistent with PANK as the feedback-regulated, rate-controlling step in the pathway.

PZ-2891 increases intracellular CoA in cultured cells.Asa prelude to examining the pantazine effect in animals, a series of PZ-2891 elevates tissue CoA content in mice. The favorable experiments were performed to verify that PZ-2891 targeted in vitro ADME and in vivo pharmacokinetic properties of PZ- PANK3 in cells and to determine if there were any obvious off- 2891 (Supplementary Tables 1, 2) suggested it could be utilized as target contraindications. A cellular thermal shift assay (CETSA)38 a proof of principle pantazine to alter CoA levels in a mouse was used to confirm that PZ-2891 was bound to cellular PANK3. model. Mice were treated with PZ-2891 either with or without a HEK 293 T cells transfected with a plasmid expressing PANK3 pantothenate supplement included in the oral gavage to deter- were used to increase cellular levels of PANK3 so that the protein mine if pantothenate administration augmented the effect of PZ- could be detected by immunoblotting cell lysates (Fig. 4a). The 2891 in mouse liver as it did in cultured cells. Treatment of mice CETSA assay showed that PZ-2891 stabilized PANK3 by ∼7°C with the maximum soluble dose of PZ-2891 (30 mg/kg) increased confirming that PZ-2891 bound to PANK3 in intact cells total liver CoA by 50% in animals administered PZ-2891 alone. (Fig. 4b). A PZ-2891 CETSA dose-response curve indicated that There was a 100% increase in CoA levels in animals treated with half of the cellular PANK3 was bound at a concentration of ∼9 PZ-2891 plus pantothenate (Fig. 5b). We next analyzed target μM PZ-2891 (Supplementary Fig. 5). PZ-2891 (10 μM) did not tissue pantothenate levels to determine the pantothenate con- impact cell viability as measured by cell growth or the synthesis of centrations in animals maintained on normal chow (25 ppm DNA, protein and lipid, indicating the absence of significant pantothenate) or animals treated with 200 mg/kg of pantothenate toxicity toward cultured cells (Supplementary Fig. 6). The pro- by oral gavage. Pantothenate supplementation increased pan- pensity of PZ-2891 to interfere with the activity of other cellular tothenate levels in plasma (Fig. 5c), liver (Fig. 5d), forebrain processes was evaluated by two enzymatic screens. First, PZ-2891 (Fig. 5e) and hindbrain (Fig. 5f). The amounts of pantothenate in did not have significant inhibitory activity in a screen of 468 liver correlated with the plasma levels, and calculation of the mammalian kinases (Supplementary Fig. 7, Supplementary intracellular concentrations in liver showed higher levels of Dataset 1). The second screen examined the effect of PZ-2891 on pantothenate than in the plasma in both cases (10.6, and 24.4 μM, 72 cellular proteins known to cause off-target effects in drug respectively). Similarly, pantothenate levels in the hindbrain and discovery (Supplementary Dataset 2). These data show that PZ- forebrain were elevated by pantothenate supplementation; how- 2891 selectively targeted PANK in cells and indicated the absence ever, the pantothenate levels in brain were significantly higher of off-target pharmacological interactions. than in liver. In all circumstances, estimated tissue pantothenate The total intracellular CoA levels were determined in a human concentrations were below the PANK3 pantothenate Km (45 liver-derived cell line (C3A) treated for 24 h with increasing μM), explaining how the pantothenate supplement potentiates concentrations of PZ-2891 (Fig. 4c). CoA progressively increased the effect of PZ-2891 on tissue CoA levels. up to 10 μM PZ-2891, whereas 20 and 50 μM were less activating. The potential therapeutic benefit of a treatment for PKAN As expected, cells did not respond to the inactive pantazine, PZ- neurodegeneration rests on its ability to cross the blood–brain 3067 (Fig. 4c). The activation phenomenon was not due to barrier and increase CoA levels in the brain. Male and female enhanced transcription of the kinase isoforms (Supplementary mice were co-administered increasing PZ-2891 concentrations to Fig. 8). To verify that PZ-2891 action was PANK-dependent, cells determine if PZ-2891 was capable of elevating CoA in the brain were transfected with an expression plasmid to increase the (Fig. 6). The animals received 5 doses of PZ-2891 by oral gavage cellular content of PANK3. The addition of 10 μM PZ-2891 to every 12 h, and tissues were harvested 4 h after the last dose. We control (empty vector) HEK293T cells also significantly increased observed a dose-dependent increase in total CoA in liver (Fig. 6a, the levels of intracellular CoA (Fig. 4d). The plasmid-driven d), forebrain (Fig. 6b, e) and hindbrain (Fig. 6c, f) in PZ-2891- increase in PANK3 expression resulted in higher baseline CoA, treated male and female animals. PZ-2891 increased CoA in both and the treatment of PANK3 overexpressing cells with 10 μM PZ- males and females to a similar extent. In this short treatment 2891 triggered a large increase in intracellular CoA (Fig. 4d). regimen, higher PZ-2891 doses were required to maximally Again, PZ-3067 was inactive. Similar transfection experiments increase CoA in the brain compared to the liver. We also showed that PANK1 expression also raised CoA after a 24 h measured PZ-2891 levels in tissues from mice treated with PZ- treatment with 10 μM PZ-2891 (Fig. 4e). There was no PZ-2891- 2891 for four weeks, and these data clearly show the presence of stimulated CoA synthesis in cells expressing the inactive PANK3 drug in liver and brain (Supplementary Fig. 11). To ensure that (E138A) mutant, consistent with a requirement for catalytically an active PANK2 was not necessary for the PZ-2891 effect, we active PANK3 rather than some other function of the protein9 treated Pank2−/− mice with PZ-2891 and measured the CoA (Fig. 4e). Along with CoA, the levels of phosphopantetheine and levels in liver and brain. PZ-2891 effectively increased CoA in the dephospho-CoA and other pathway intermediates were measured tissues of Pank2−/− knockout mice (Supplementary Fig. 12) (Supplementary Fig. 9). These two pathway intermediates were illustrating that the expression of PanK2 was not required for minor components in both control and PZ-2891-treated cells pantazine activation of CoA biosynthesis, and that activation of showing that activation of CoA biosynthesis at the PANK step PanK1 and PanK3 were sufficient to raise CoA. PZ-2891 therapy was not constrained by a secondary control point in the CoA would be maintained for extended periods, therefore we biosynthetic pathway. Mass spectrometry showed that PZ-2891- formulated pantothenate-supplemented mouse chow with differ- treated cells had higher acetyl-CoA levels compared to untreated ent concentrations of PZ-2891. Five mice were treated for 1 week controls (Supplementary Fig. 10), indicating that PZ-2891 with 112 ppm PZ-2891 in the chow, and the elevation in CoA in effectively prevented feedback inhibition by acetyl-CoA. Once six tissues were examined (Supplementary Fig. 13). CoA levels PANK3 is in the active state, the reaction rate is determined by were elevated in liver, brain, and to a lesser extent, heart. Muscle the concentration of pantothenate in relation to the PANK Km CoA levels are the lowest of all tissues39, and were not elevated by

6 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE

a Control 55 57 59 61 63 65 67 °C

PANK3

Actin

PZ-2891

PANK3

Actin

bc2.0 Control 100 PZ-2891 PZ-2891 1.5 cells 7

1.0 50 Percent of control Percent nmol CoA/10 0.5 PZ-3067

0 0.0 50 55 60 65 70 75 0 10203040 50 Temperature (°C) Compound (μM)

de30 PANK1β 40 PANK3 pCDNA3.1 PANK3(E138A) PANK3 30 20 cells 7 cells

7 20

10 4 nmol CoA/10

nmol CoA/10 2 0 Control PZ-2981 0 DMSO PZ-2891 PZ-3067 α-PANK 38 kDa

α-GAPDH 37 kDa

Fig. 4 Pantazine modulation of intracellular CoA levels. a Representative immunoblots from a cellular thermal shift (CETSA) assay showing the stabilization of PANK3 by PZ-2891. b PANK3 stability in the presence (red open circles) and absence (filled black circles) of 10 μM PZ-2891. These data were obtained from three independent biological experiments. c Intracellular CoA levels in C3A cells. Cell cultures were treated with either 10 μM PZ-2891 (red open circles) or its inactive homolog, PZ-3067 (filled black circles), and the CoA levels were determined. These data were derived from three independent biological experiments. d HEK293T cells were transfected with either an empty vector (gray bars) or vector driving the expression of PANK3 (pink bars), and the cells were treated with either 10 μM PZ-2891 or PZ-3067. The data are from two independent biological experiments performed in duplicate and the mean ± SEM is plotted. e HEK293T cells were transfected with expression vectors for either PANK1β (white bars), PANK3 (pink bars), or PANK3 (E138A) (orange bars) . PANK3(E138A) is a catalytically inactive mutant. Triplicate dishes of cells were treated for 24 h with either DMSO or 10 μMPZ- 2891 in DMSO, the cells were harvested and the total intracellular CoA levels determined. Data were plotted as the mean ± SEM. Isoform expression was verified by immunoblotting cells samples with an anti-PANK antibody (α-PANK) raised against the catalytic core common to all isoforms

PZ-2891 therapy. A dose-response experiment was conducted by manner. During the one-month treatment there was no difference the administration of a series of PZ-2891 doses in the diet in the weights or behavior of the treated animals. These data (Fig. 6g, i). These experiments showed both liver and brain CoA established that long-term PZ-2891 therapy was effective in were significantly elevated by PZ-2891 in a dose-dependent elevating CoA in the liver and brain of mice.

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2

‘fl ’ ab5 250 oxed alleles, and the Cre recombinase controlled by the Control <0.0001 synapsin I promoter (SynCre) transgene gives rise to neuron- PZ-2891 selective deletion of both alleles (Supplementary Fig. 14). Control 4 200 <0.0001 animals were littermates lacking the SynCre transgene. The brains +

cells of SynCre animals analyzed by RT-PCR showed the reduction

7 3 150 in both Pank1 and Pank2 mRNA, but not Pank3, in brain (Supplementary Fig. 15a). The same analysis on liver tissue 2 100 showed that all three Pank genes were normally expressed, Liver total CoA Liver illustrating the selective deletion of Pank1/Pank2 in the brain nmol CoA/10 1 weight) (pmol/mg wet 50 (Supplementary Fig. 15b). The mice develop normally until postnatal day 12.5, at which time they begin to lose weight 0 0 (Fig. 7a), and exhibit a median survival of 52 days (Fig. 7b). PZ- Pan – + – + 0.04 0.4 4 2891 (112 ppm in chow) treatment began at weaning (day 21) μ PZ-2891 – – + + Pantothenate ( M) after the disease was established and significant weight loss had occurred. The treated animals immediately began to gain weight c 20 d 30 after therapy was initiated (Fig. 7a), and had a median survival of <0.0001 <0.0001 150 days (Fig. 7b). Two of 11 treated animals were so affected that M)

μ they died within two days of starting PZ-2891, and 5 of 11 treated 15 20 animals lived until the study was terminated at six months. Day 45 was selected as a time to compare treated and untreated ani- 10 mals (Fig. 7a). The SynCre+ animals had lower CoA in the brain compared to their littermate controls, and PZ-2891 therapy sig- 10 nificantly elevated CoA in both the forebrain and hindbrain

5 pantothenate Liver + fi (pmol/mg wet weight) (pmol/mg wet (Fig. 7c, d). The SynCre animals exhibited a signi cant loco-

Plasma pantothenate ( motor defect as measured in the open field test, an accepted 0 0 method for assessing mouse neuromuscular disorders. The 0 200 0 200 movement of SynCre+ animals was severely impaired (Fig. 7e), Pan (mg/kg) Pan (mg/kg) and when they did move, they did not travel far (Fig. 7f). PZ-2891 treatment of the SynCre+ animals significantly increased loco- ef150 motor activity (Fig. 7e,f; Supplementary Fig. 16). PZ-2891 did not 150 <0.0001 0.0022 affect control mouse locomotor activity. These data show that PZ- 2891 therapy increases weight, life span, and brain CoA levels, as well as improving the locomotor activity of animals with severe 100 100 growth and locomotor defects due to a deficiency of pantothenate kinase activity and CoA in neurons.

50 50 Discussion (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet Hindbrain pantothenate Hindbrain Forebrain pantothenate Forebrain PKAN is a progressive neurodegenerative disorder affecting movement, balance, speech, vision, cognition and behavior, and it 0 0 4 0 200 0 200 arises from debilitating mutations in the PANK2 gene .PANK2isa Pan (mg/kg) Pan (mg/kg) major PANK isoform expressed in the brain, and PKAN symptoms are thought to arise from a CoA deficiency that compromises Fig. 5 Pantothenate concentration and PZ-2891 activation of CoA synthesis important neuronal processes including iron metabolism, synaptic in animals. a Pantothenate-dependent PZ-2891 elevation of CoA in cultured transmission, synaptic vesicle cycling, neuron projection develop- C3A cells. Pantothenate levels were manipulated by supplementing ment, and protein quality control40. There is no effective PKAN pantothenate-free DMEM + dialyzed FBS. Control CoA levels (gray bars) treatment. Our therapeutic approach is to compensate for the loss are compared to cells treated with 10 μM PZ-2891 (pink bars). Data are of PANK2 by activating the two other PANK isoforms using a small from two experiments performed in duplicate, and plotted as the mean ± molecule (PZ-2891) that penetrates the blood brain barrier to ele- SEM. b Groups of 5 mice were treated with five doses of PZ-2891 (30 mg/ vate tissue CoA. PZ-2891 is the result of a successful selection/ kg) without or with 200 mg/kg pantothenate. PZ-2891 was administered at optimization process designed to exploit the high cooperativity 12 h intervals, and 4 h after the last dose the animals were euthanized and between PANK protomers that simultaneously switch from an the CoA content of the tissues determined. Untreated mice (filled circles); active ATP-bound state to an inactive acetyl-CoA-bound state. PZ- pantothenate-treated mice (filled squares), PZ-2891 treatement (open 2891 exploits this mechanism by engaging PANK across the dimer circles), and PZ-2891 + pantothenate (open squares). c–f Groups of 5 interface to act as an orthosteric inhibitor at high concentrations animals were gavaged with 200 mg/kg of pantothenate using the same and an allosteric activator at lower sub-saturating concentrations. regimen as in b, and four hours after the last dose, tissues were harvested Thus, PZ-2891 relaxes the feedback inhibition of PANK resulting in and the pantothenate concentrations measured by mass spectrometry. asignificant rise in CoA levels in cells and tissues. Due to the c Plasma. d Liver. e Forebrain. f Hindbrain. Mean ± SEM are plotted, and the neuronal nature of PKAN, the design process incorporated physi- p values are in red. Control pantothenate levels (filled circles), cochemical properties that reflect the top 25 drugs which target the pantothenate levels in pantothenate-treated mice (open circles) central nervous system31,32, resulting in blood brain barrier pene- tration and the elevation of brain CoA levels. A mouse model of PZ-2891 therapy increases lifespan in a mouse model of brain neuronal CoA deficiency was developed to evaluate the therapeutic fi CoA de ciency. We developed a mouse model for neuronal CoA potential of the pantazines. These animals have reduced CoA in the fi de ciency as a platform to test therapeutics (Fig. 7). These ani- brain, fail to gain weight, have a severe locomotion defect, and a mals have both the Pank1 and Pank2 genes as homozygous reduced life span. PZ-2891 treatment elevates brain CoA, the

8 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE

a PO MALE dgPO FEMALE DIET MALE 250 250 250 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0002 200 200 0.001 200 <0.0001 0.0004 0.0036 150 150 150

100 100 100 Liver total CoA Liver Liver total CoA Liver Liver total CoA Liver (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet 50 50 50

0 0 0 0 5 10 15 20 0 5 10 15 20 0 7.5 37.5 75 225 PZ-2891 (mg/kg) PZ-2891 (mg/kg) PZ-2891 (ppm)

be120 120 h120

<0.0001 0.0004 90 0.0026 90 90 <0.0001 <0.0001 0.0003 0.0012 60 60 60

30 30 30 Forebrain total CoA Forebrain Forebrain total CoA Forebrain Forebrain total CoA Forebrain (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet

0 0 0 0 5 10 15 20 0 5 10 15 20 0 7.5 37.5 75 225 PZ-2891 (mg/kg) PZ-2891 (mg/kg) PZ-2891 (ppm)

c 150 f 150 i 150 <0.0001 <0.0001 0.0123 <0.0001 0.0002 120 120 0.0014 120 0.0009 0.0008 90 90 90

60 60 60 Hindbrain total CoA Hindbrain Hindbrain total CoA Hindbrain Hindbrain total CoA Hindbrain (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet 30 weight) (pmol/mg wet 30 30

0 0 0 0 5 10 15 20 0 5 10 15 20 0 7.5 37.5 75 225 PZ-2891 (mg/kg) PZ-2891 (mg/kg) PZ-2891 (ppm)

Fig. 6 Activation of CoA synthesis in liver, forebrain and hindbrain of male and female animals treated with PZ-2891. Groups of 5 mice were administered the indicated amounts of PZ-2891 plus 100 mg/kg pantothenate by oral gavage every 12 h for 5 doses. Four hours after the last dose, the tissues were harvested and the tissue CoA levels determined. Control CoA levels (closed circles); PZ-2891 treated CoA levels (open circles). a Male liver. b Male forebrain. c Male hindbrain. d Female liver. e Female forebrain. f Female hindbrain. Groups of 5 mice were maintained on chow fortified with 1000 ppm pantothenate and the indicated levels of PZ-2891. The mice were maintained on the diets for 4 weeks and their tissues were harvested and the CoA levels determined. g Liver CoA levels. h Forebrain CoA levels. i Hindbrain CoA levels. Statistical significance was determined using Student’s t-test calculated with Graph-Pad software and the p values (red) are noted on the figure panels in red. Means ± SEM are plotted treated animals gain weight, their lifespan increases significantly, the cellular effects of PZ-2891 verify that pantothenate kinase is and they exhibit increased locomotor activity. These data support the only relevant rate-controlling enzyme in mammalian CoA the conclusion that the activation of alternate pantothenate kinase biosynthesis. Although CoA synthase was suggested as a second isoforms by pantazines is a promising approach to PKAN therapy. control point in the CoA biosynthetic pathway6,43–45, our The pantazines are also powerful chemical probes to inter- quantitative identification of pathway intermediates by mass rogate the role of CoA homeostasis in cell physiology. The bio- spectrometry shows that significant activation of PANK does not chemical mechanism of activation of PZ-2891 is different from result in the accumulation of other pathway intermediates, such other previously described allosteric kinase activators that act by as phosphopantetheine, the CoA synthase substrate. The steady either destabilizing the inactive kinase conformation or stabilizing state CoA tissue level is controlled by the relative rates of the active one41,42. This unique mechanism of action makes PZ- synthesis and turnover. CoA turnover is thought to be catalyzed 2891 a highly specific molecule that does not significantly interact by nudix hydrolases11,46, but in cultured cells, nudix hydrolase with other mammalian kinases or receptors. Another attractive expression is low and CoA turnover may be slower than in animal feature of PZ-2891 is its relatively long residence time on PANK. tissues47. Current medicinal chemistry efforts are directed toward Residence time on the cellular target is a key parameter in gov- optimizing the pharmaceutical properties of PZ-2891 while erning drug efficacy37, and in the case of PZ-2891, mitigates its maintaining its excellent oral bioavailability, target affinity, and relatively fast metabolic clearance from the body. Our studies on blood brain barrier permeability.

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2

a 35 b PZ-2891 diet 100 PZ-2891 diet 30 Regular diet (n = 11) p = 0.0008 25 Sample 20

15 Start of 50 treatment Body weight (g) Body weight 10 survival Percent Regular diet (n = 11) 5

0 0 0 50 100 150 200 0 50 100 150 200 Postnatal days Postnatal days

cd80 150 <0.0001 <0.0001 <0.0001 <0.0001 60 100 (n =5) (n = 10) (n = 10) (n =5) 40 (n =7) 50 (n =7) Hindbrain total CoA Hindbrain Forebrain total CoA Forebrain (pmol/mg wet weight) (pmol/mg wet (pmol/mg wet weight) (pmol/mg wet 20

0 0 SynCre – + + SynCre –++ PZ-2891 ––+ PZ-2891 ––+

ef100 25

<0.0001 80 20

60 15 <0.0001 (n = 10) (n = 12)

40 10 (n = 10) (n = 12) %Time moving (n = 11) (m) traveled Path 20 (n =8) 5 (n = 11) (n =8) 0 0 SynCre ––+ + SynCre ––+ + PZ-2891 – + – + PZ-2891 – + – +

Fig. 7 PZ-2891 therapy in SynCre+ PANK1,PANK2 neuronal knockout mice. Male and female mice were randomized into the experimental groups when they were genotyped, and the numbers of mice used in each analysis are shown in parenthesis in the figure panels. a Weight monitoring of double knockout mice maintained on the control diet (black circles) or on a diet with PZ-2891 (red circles). The start of treatment and the time of sample collection are indicated with arrows on the figure. b Lifespan of double knockout mice with (red line) or without (black line) PZ-2891 treatment. c Forebrain CoA levels in littermate control (SynCre−) mice (filled circles) compared to SynCre+ Pank1,Pank2 neuronal knockout mice at day 45 with (open squares) or without (filled squares) PZ-2891 therapy. d Hindbrain CoA levels in littermate control (SynCre−) mice (filled circles) compared to SynCre+ Pank1,Pank2 neuronal knockout mice at day 45 with (open squares) or without (filled squares) PZ-2891 therapy. e The percent time mice were moving during the 5 min open field test. Littermate control (SynCre−) mice (untreated, filled circles; treated open circles) were compared to SynCre+ Pank1,Pank2 neuronal knockout mice (untreated filled squares; treated, open squares) at day 45 with or without PZ-2891 therapy. f Distance traveled by mice in the 5 min open field test. Littermate control (SynCre−) mice (untreated, filled circles; treated, open circles) were compared to SynCre+ Pank1,Pank2 neuronal knockout mice at day 45 with (open squares) or without (closed squares) PZ-2891 therapy. The p values are in red, and means ± SEM are plotted

Methods Dynamic-injection surface plasmon resonance. Experiments were conducted at Materials. D-[1-14C]Pantothenate (specific activity, 55 mCi/mmol) from 20 °C using a SensiQ Pioneer optical biosensor (SensiQ Technologies). His-tagged American Radiolabeled Chemicals; Ni-NTA resin from Qiagen; [3H]ATP from human PANK3 was immobilized on polycarboxylate hydrogel-coated gold chips Perkin Elmer (specific activity, 29.8 Ci/mmol); Sypro Orange dye from Thermo preimmobilized with nitrilotriacetic acid (HisCap chips; SensiQ Technologies). The Fisher Scientific; D-[3H]pantothenate (specificactivity,50Ci/mmol)from chip was primed in chelating buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 50 μM American Radiolabeled Chemicals; D-pantothenic acid, hemicalcium salt from EDTA, 0.005% Tween-20) and was preconditioned at 10 μl/min with three 60 s Sigma-Aldrich; pantothenate-free DMEM from Thermo Fisher Scientific; CoA injections of wash buffer (10 mM HEPES, pH 8.3, 150 mM NaCl, 350 mM EDTA, thioesters from Avanti Polar Lipids. All other materials were reagent grade or 0.05% Tween-20) and one 60 s injection of chelating buffer before being charged μ better. with a 60 s injection of 500 M NiCl2 in chelating buffer. After priming into

10 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE binding buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1 mM TCEP, 2 mM fetal bovine serum (FCS, Altanta Biologicals), 50 U/ml penicillin and 50 mg/ml fi MgCl2, 1 mM ATP, 0.005% Tween-20, 3% DMSO), PANK3 was injected until ~ streptomycin. Both cell lines were con rmed to be mycoplasma-free. 3000–4000 RU of protein were captured. The reference flow cell on the chip was HEK293T cells were used for overexpression studies and were transfected using charged with Ni2+ without adding protein. FuGENE® 6 reagent (Promega) combined with 12 μg of the indicated plasmid PZ-2891 was prepared in binding buffer, and a gradient of 0–50 or 0–100 nM cDNA construct: pcDNA3.1(−) vectors expressing either human PANK3, was injected in duplicate for each concentration at a flow rate of 200 μl/min using PANK1β, PANK2m or PANK3(E138A) according to the manufacturer’s recom- the OneStep Injection feature, which exploits Taylor dispersion to generate a mendations. PZ-2891, PZ-3067 or vehicle control was added 24 h post-transfection. concentration gradient that provides a full titration of analyte in a single injection. After 24 h of treatment (total 48 h after transfection) the cells were washed with A series of buffer-only (blank) injections was included to account for instrument PBS and harvested and subjected to either total CoA or acetyl-CoA determinations. noise. PANK3 was released from the chip with 500 mM imidazole and recaptured For radiolabeling human C3A cells with pantothenate, medium was prepared for each cycle due to the slow dissociation of the compound. The data were from glucose-, pantothenate-, pyruvate-, choline chloride-, methionine- and processed, double-referenced, solvent corrected and analyzed using the software glutamine-free DMEM (Gibco/Life Technologies/ThermoFisher) reconstituted package Qdat (version 2.5.3.5, SensiQ Technologies). Kinetic rate constants (ka, kd) with 1 g/l glucose, 1 mM pyruvate, 1 mg/ml choline chloride, 15 mg/ml methionine, were determined by globally fitting the data at all concentrations to a 1:1 2 mM glutamine, and 1 mCi/ml [2,3–3H]pantothenic acid (specific activity 50 Ci/ (Langmuir) binding model. Equilibrium dissociation constants were calculated as mmol; American Radiolabeled Chemicals, ART 0731), 10% delipidated FBS = – the quotient KD kd/ka. (Cocalico Biologicals #55 0115), 50 U/ml penicillin, and 50 mg/ml streptomycin. Prior to labeling, cells were seeded at a density of 2–3×106 in 60 mm-dishes in reconstituted DMEM without added pantothenate and allowed to adhere PANK activity assays . PANK activity assay was performed in the presence of overnight. The labeling was initiated by replacing the medium with 2 ml of labeling 0–10 μM compound in a reaction mixture that contained 100 mM Tris-HCl, pH μ μ –14 fi medium containing 0.1% dimethylsulfoxide (DMSO; control) or 10 M of the 7.5, 10 mM MgCl2, 2.5 mM ATP, 45 M D-[1 C]pantothenate (speci c activity, compound in 0.1% DMSO. Cells were incubated for 24 h, the labeling medium was 22.5 mCi/mmol) and 5 nM of human PANK3. PANK3 concentrations were cal- fi −1 −1 removed, and cells were washed with phosphate buffered saline and lysed by culated using the extinction coef cient at 280 nm of 39,225 M cm . The assay sonication in 20 mM Tris-HCl, pH 7.5, 2 mM DTT, 5 mM EDTA and 50 mM NaF. was linear with time and after 10 min at 37 °C the reaction was stopped by the μ The lysates were fractionated by thin-layer chromatography after removal of the addition of 4 l of 10% (v/v) acetic acid. The mixture was spotted onto a DE81 disk, cell debris by centrifugation11. The samples were spotted on Silica Gel H plates and washed with three successive changes of 1% acetic acid in 95% ethanol and product 33 resolved with 95% ethanol:28% ammonium hydroxide (4/1, v/v). The distribution formation determined by scintillation counting of the dried disc . If the IC50 was of radioactivity on the plate was quantified using a Bioscan Imager (Bioscan, Inc.) determined to be in the nM range then the assay was repeated in the presence of fi – μ – μ together with quanti cation of the total cellular radioactivity determined by a 0 1 Mor0 0.1 M compound to more precisely determine the IC50. All the Liquid Scintillation Analyzer (PerkinElmer Tri-Carb 2910 TR). Triplicate culture experiments were repeated twice in duplicate and the data were an average ± data dishes were labeled for each group plus a dish for each group for determination of range. For the kinetic experiments, the assay was done either varying the pan- fi – μ – μ cell counts and viability using a Nucleocounter (New Brunswick Scienti c) tothenate from 0 180 M or ATP from 0 125 M at a given concentration of PZ- according to the manufacturer’s directions. The experiments were repeated twice in 2724. duplicate and the data represented is an overall average ± SEM. The experiments mimicking the mixture of ligands present in vivo were For quantifying protein, DNA and lipid synthesis, C3A cells were first seeded at performed under different conditions. The reaction mix for the determination of 2×106 in 60 mm-dishes and grown overnight in maintenance medium. Cells were the acetyl-CoA IC50 contained 100 mM Tris-HCl (pH 7.5), 10 mM MgCl2,1mM μ 14 fi μ washed once with phosphate buffered saline (PBS) and 2 ml of labeling medium ATP, 45 M D-[1- C]pantothenate (speci c activity 22.5 mCi/mmol), 2.5 M PZ- plus radiochemical was added to start the 24 h incubation. Labeling medium 2891 and 1 μg of PANK3. In the time course experiments, the reaction mixtures μ μ consisted of 90% DMEM lacking glucose and pyruvate (Life Technologies), 10% contained 100 mM Tris-HCl (pH 7.5), 10 mM MgCl2, 1 mM ATP, 45 Mor90 M –14 fi μ DMEM containing 4.5 g/l glucose, lacking glutamine (Biowhittaker/Lonza) plus D-[1 C]pantothenate (speci c activity, 22.5 mCi/mmol), 100 M acetyl CoA ± 10% delipidated FBS (Cocalico Biologicals). A final concentration of 5 μCi/ml 3H- 2.5 μM PZ-2891 and 1 μg of PANK3. fi – fi labeled amino acid mixture (speci c activity 60 Ci/mmol; American Radiolabeled The IC50 values for the structure activity study were calculated by tting the μ 3 fi – Chemicals, ART 0328), 5 Ci/ml [ H]thymidine (speci c activity 84 Ci/mmol; inhibition data to a one-site model of the Michaelis Menten equation. Although American Radiolabeled Chemicals, ART 0178) or 5 μCi/ml [U-14C]glucose this method was appropriate to measure the IC50 with most of the pantazines, it fi fi (speci c activity 10 mCi/mmol; American Radiolabeled Chemicals, ART 0122 A) underestimates ligand binding af nity if the concentration of protein in the assay was used to measure protein synthesis, DNA synthesis or lipid synthesis, alters the free ligand concentration. Thus, the PZ-2891 data were fit to Morrison’s 48 respectively. The labeling medium contained 0.1% dimethylsulfoxide (DMSO; quadratic equation (GraphPad software) that accounts for the impact of enzyme- control) or 10 μM compound in 0.1% DMSO. Cells were harvested, washed once inhibitor binding on the free concentration of inhibitor. with PBS and harvested either by trypsinization and vacuum filtration through 0.45 micron discs (Millipore, HAWP02500) for protein and DNA determinations, μ + μ + Protein thermal shifts. Protein thermal shift experiments were conducted in a or by scraping and extraction in 100 lH2O 240 l methanol/acetic acid 98:2 μ mixture containing 100 mM HEPES, pH 7.0, 5X Sypro Orange dye, 10 mM MgCl2, 100 l CHCl3 for lipid determination. Filters were washed three times with PBS, air 2.5 μM PANK3 and ATP, acetyl CoA or pantazines at the indicated concentrations. dried and counted with a Liquid Scintillation Analyzer. Radioactivity was Reaction mixture (100 μl) was aliquoted into optically clear 96-well PCR plates, normalized to cell numbers and viability was >95%. Cellular lipid emulsions were centrifuged and placed in an ABI 7500 real time PCR system for thermal shift separated by centrifugation and the lower phases were quantified by liquid analysis. The temperature was ramped from 25 °C to 95 °C at a rate of 1 °C/min, scintillation counting and normalized to cell numbers. and fluorescence was measured using a TAMRA filter set (Ex 560 nm and Em 582 nm). The data were plotted as fluorescence intensity as a function of temperature Human PANK isoform distribution. The levels of human PANK mRNA were and were fit to the first derivative of the Boltzman sigmoidal equation using determined by real time qPCR in C3A cells that were grown to a density of 3 × 106 GraphPad software to determine the temperature corresponding to the denatura- cells/dish and were treated either with 10 μM PZ-2891 or DMSO for 24 h. RNA tion of 50% of the protein. Each experiment was performed in triplicate and the was isolated immediately using Trizol reagent according to the manufacturer’s melting temperatures were the average of 3 independent data sets rounded to the instructions (Invitrogen). The cDNA was synthesized using Super ScriptTM II, nearest degree. For determining the K , the change in melting temperature from 0.5 random primers and the RNA templates after the removal of genomic DNA by all 3 data sets was plotted against concentration of the ligand, and those data were turbo DNAse (Ambion). Quantitative real-time PCR was performed using SYBR fit to the Michaelis–Menton equation using GraphPad software. Green master mix in triplicate with the primers listed below. The human glyceraldehyde-3-phosphate dehydrogenase GAPDH (Applied Biosystems) was 3 Gel filtration chromatography.[H]ATP (300 μM; 1.38 Ci/mmol) was incubated used as a control. All of the real-time values were compared using the CT method, with PANK3 in a mixture containing 100 mM Tris pH 7.5, 10 mM MgCl2, with or where the amount of PANK1, 2 or 3 cDNA was normalized to the housekeeping Δ without 5 μM PZ-2891 on ice for 10 min, and then applied to a Superdex 75 10 gene GAPDH ( CT) before being compared with the amount of PANK cDNA ΔΔ 300GL column (GE Lifesciences) equilibrated and eluted with 20 mM Tris-HCl, pH under DMSO condition ( CT), which was set as the calibrator at 1.0. The data 7.5, 200 mM NaCl. Fractions (200 μl) were collected, and 100 μl were counted by represent an average of three experiments. The primers are listed in Supplementary scintillation counting. The radioactivity counts obtained were then plotted against Table 5. the elution volume and compared to the absorbance values obtained at 280 nm in the elution profile. The gel filtration experiment was performed twice, and one of PANK antibody the experiments is represented here. . The pan-reactive anti-PANK antibody that recognized all mouse and human PANK isoforms was purified from rabbit polyclonal antiserum that was raised against pure full-length recombinant human PANK1β. Immunization of Cell culture and radiolabeling. HEK293T cells (ATCC, #CRL-3216) were main- rabbits and collection of antisera were performed by Rockland Immunochemicals, tained in Dulbecco’s modified Eagle’s medium (DMEM, Biowhittaker/Lonza) Inc. according to their standard schedule. Antisera were purified by affinity supplemented with 10% FCS. Human C3A [HepG2/C3A, derivative of HepG2] chromatography on Affi-Gel cross-linked with the recombinant PanK1β protein. cells (ATCC #CRL-10741) were purchased from ATCC® and maintained in Eagle’s The purified α-PANK antibody was validated by immunoblotting serial dilutions of minimum Essential medium (ATCC) supplemented with 2 mM glutamine, 10% the purified antibody preparations against purified recombinant mouse and human

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2

PANK1α, PANK1β, PANK2 and PANK3 proteins, and cell lysates expressing each min to pellet debris, and the supernatant was transferred to a glass vial. A PZ-2891 of the isoforms (Supplementary Fig. 17). The entire blots for the cropped data in standard curve was created by spiking in known concentrations of PZ-2891 into 20 Fig. 4a and Fig. 4e are presented as Supplementary Fig. 18 and Supplementary μl of plasma from a control mouse and following the above procedure. Tissue was Fig. 19, respectively. homogenized in 2 ml of 80% methanol containing 0.1 μM warfarin, and incubated at −80 °C for 4 h. Samples were spun at 3500×g for 10 min to pellet debris, supernatant was transferred to a glass tube and dried down using a Savant Cellular thermal shift. The method was adapted from Jafari, et al38. SPD1010 Speed-Vac (Thermo Scientific) overnight. Samples were resuspended in HEK293T cells were transfected with 5 μg/dish pPANK3, a pcDNA3.1(−)-derived 400 μl of 80% acetonitrile to a final concentration of 0.5 μM and transferred to a expression vector, in a 100 mm dish. After 24 h the cells were treated with the glass vial. indicated concentrations of pantazine or 0.1% DMSO for 2 h at 37 °C. Four dishes PZ-2891 was analyzed using a Shimadzu Prominence UFLC attached to a Sciex were used in each assay. The cells were washed and resuspended in PBS, counted to QTrap 4500 equipped with a Turbo V ion source. Samples (5 μl) were injected onto determine viable cells and then were resuspended in PBS at 5 × 105 cells/100 μl. an XSelect® HSS C18, 2.5 μm, 3.0 × 150 mm column (Waters) using a flow rate of Aliquots (100 μl) in 0.5 ml reaction tubes were exposed to various temperatures 0.25 ml/min. Solvent A was 0.1% formic acid in water, and Solvent B was (25, 53, 55, 57, 59, 61, 63, 65 and 67 °C) in pairs (control and compound treated) acetonitrile + 0.1% formic acid. The HPLC program was the following: starting using a thermo cycler PCR machine for 3 min and then moved to room tem- solvent mixture of 50% B, 0–0.5 min isocratic with 50% B; 0.5–1.5 min linear perature for another 3 min before being flash frozen in liquid nitrogen, and stored gradient to 95% B; 1.5–20 min isocratic with 95% B; 20–21 min linear gradient to at −80 °C. The samples were freeze-thawed 5 times, sonicated at 4 °C for 10 s, and 50% B; 21–25 min isocratic with 50% B. The QTrap 4500 was operated in the the lysates were centrifuged at 14,000×g for 20 min to remove precipitated proteins. positive mode, and the ion source parameters were: ion spray voltage, 5500 V; The soluble fractions were fractionated on a 10% bis-Tris acrylamide gel and curtain gas, 30 psi; temperature, 450 °C; collision gas, medium; ion source gas 1, 30 transferred to a PVDF membrane. The blots were blocked for 1 h in 1% milk/TBS- psi; and ion source gas 2, 40 psi. The MRM transition for PZ-2891 was 350.2 / T and then exposed to primary antibody (α-PANK) overnight at 1 μg/ml in 1% 190.0 m/z and warfarin was 309.1 / 163.0 m/z both with a declustering potential, 65 BSA/TBS-T followed by secondary antibody (anti-Rabbit AP conjugated) in 1% V and collision energy, 30 V. The system was controlled by the Analyst® software milk/TBS-T for 1 h at 1:5000 dilution. The blot was washed extensively and (Sciex) and analyzed with MultiQuant™ 3.0.2 software (Sciex). exposed to the ECF substrate for 5 min, and the bands on the dried membrane were quantified on the Typhoon FLA9500 using ImageQuant TL software (GE Healthcare). The control blot was done with samples exposed to different tem- Pantothenate extraction and quantification by HPLC/MS/MS. Plasma (10 μl) peratures but were not centrifuged to remove the precipitated proteins and probed was added to 800 μl methanol and 190 μl water along with 200 pmol (β-alanyl- β 13 15 13 15 – with anti- -actin antibody (Sigma-Aldrich). All the experiments were repeated C3 N)-pantothenate ( C N-Pan, Sigma). Tissue (30 40 mg) was homogenized twice in duplicate and the combined data are presented (overall average ± SEM). in 2 ml of 80% methanol, and 500 pmol of 13C15N-Pan was added. The samples were incubated at −80 °C for 4 h. Samples were spun at 3500xg for 10 min to pellet Tissue and cellular CoA determinations. Cultured cells or frozen tissues were debris, and supernatant was transferred to a new tube and dried using a SpeedVac μ overnight. Samples were resuspended in water + 0.1% formic acid and spun resuspended in 2 ml cold water to which 500 l of 0.25 M KOH was added, deri- μ vatized with monobromobimane (mBBr, Life technologies) and quantified by through a Spin-X Centrifuge Tube Filter (0.22 m Cellulose Acetate, Costar). HPLC23. For acetyl-CoA determinations, C3A cells were resuspended in 1 ml Pantothenate was analyzed using a Shimadzu Prominence UFLC attached to a water, added to 2 ml of methanol and 1 ml chloroform, and incubated on ice for 15 QTrap 4500 equipped with a Turbo V ion source (Sciex). Samples were injected 13 onto an XSelect® HSS C18, 2.5 μm, 2.1 × 150 mm column using a flow rate of 0.2 min. Chloroform 1.5 ml, 1.2 ml water and 30 pmol of [ C2]acetyl-CoA (Sigma) + was added, and centrifuged at 2000×g for 10 min. The top layer was loaded on a 2- ml/min. Solvent A was 0.1% formic acid in water, and Solvent B was acetonitrile (2-pyridyl) ethyl solid phase extraction column which was equilibrated with 1 ml 0.1% formic acid. The HPLC program was the following: starting solvent mixture of 1% B, 0–6 min linear gradient to 60% B; 6–13 min linear gradient to 100% B; 50% methanol/2% acetic acid. The column was washed twice with 1 ml 50% – – – methanol/2% acetic acid and 1 ml water. CoA and thioesters were eluted twice with 13 16 min isocratic with 100% B; 16 17.5 min linear gradient to 1% B; 17.5 20 1 ml 95% ethanol containing 50 mM ammonium hydroxide. Samples were resus- min isocratic with 1% B. The QTrap 4500 was operated in the positive mode, and pended in 90% methanol containing 15 mM ammonium hydroxide. the ion source parameters were: ion spray voltage, 5500 V; curtain gas, 30 psi; temperature, 400 °C; collision gas, medium; ion source gas 1, 20 psi; and ion source gas 2, 25 psi. The MRM transition for pantothenate was 220.0 / 184.1 m/z and CoA, dephosphoCoA and phosphopantetheine determinations. The mBBr [13C15N]pantothenate was 224.0 / 188.1 m/z both with a declustering potential, 35 derivatized sample was fractionated by reverse-phase HPLC using a Gemini C18 3 V and collision energy, 22 V. The system was controlled by the Analyst® software μm column (150 × 4.60 mm) from Phenomenex. The chromatography system was (Sciex) and analyzed with MultiQuant™ 3.0.2 software (Sciex). Intracellular a Waters e2695 separation module with a UV/Vis and fluorescence detector and pantothenate was measured as pmoles/mg wet weight. controlled by Empower 3 software. Solvent A was 50 mM potassium phosphate pH 4.6, and solvent B was 100% acetonitrile. Twenty microliters of sample was injected onto the column, and the flow rate was 0.5 ml/min. The HPLC program was the Animal ethics statement. All procedures were reviewed and approved by the St. ’ following: starting solvent mixture of 90% A/10% B, 0–2 min isocratic with 10% B, Jude Children s Research Hospital Institutional Animal Care and Use Committee. 2–6 min linear gradient from 10% B to 15% B, 6–18 min concave gradient from – – 15% B to 40% B, 18 23 min isocratic with 40% B, 23 25 min linear gradient from Mouse efficacy testing. C57Bl6/J mice (8-week-old) were purchased from Jackson 40 to 10%, and 25–30 min isocratic with 10% B. The UV/vis detector was set at 393 fl Laboratory and were fed either a regular diet (autoclavable rodent breeder diet, nm, and the uorescence detector was set with excitation at 393 nm and emission Labdiet) or an isocaloric, pantothenic acid-supplemented diet (1000 ppm) for at 470 nm. The elution position of the mBBr-CoA, mBBr-dephospho-CoA (deP- 2 weeks prior to the experiment. The mice were maintained at room temperature CoA), and mBBr-phosphopantetheine (PPanSH) were determined by comparison 72° ± 2 °F, humidity 50 ± 10% and a 14 h light /10 h dark cycle with the dark cycle with mBBr-CoA prepared from commercial CoA (Avanti Polar Lipids), mBBr- starting at 18:00 h. Water was supplied ad libitum. The mice were randomized into deP-CoA prepared from commercial deP-CoA (Sigma-Aldrich), and mBBr- the treatment arms to achieve a normal weight distribution. PZ-2891 was for- PPanSH prepared from a Nudt7-mediated hydrolysis of mBBr-CoA. The areas mulated in 30% Captisol49, and was administered by oral gavage at 12 h intervals under the mBBr-derivatized CoA, deP-CoA and PPanSH peaks were integrated for 5 doses. The mice were euthanized and tissues harvested 4 h after the last dose. and compared to known concentrations of the mBBr-CoA standard. The tissue samples were used for total CoA, pantothenate and pantazine deter- minations. Blood was collected from euthanized animals, plasma or serum was Acetyl-CoA measurement by mass spectrometry. Mass spectrometry of acetyl- prepared and stored frozen until analysis. Organs, including liver, forebrain, and CoA was performed using a Finnigan TSQ Quantum (Thermo Electron) triple- hindbrain were quickly excised from euthanized animals and immediately flash ® quadrupole mass spectrometer. The instrument was operated in positive mode frozen in liquid N2 or immersed in RNAlater (Qiagen) overnight prior to freezing. using single ion monitoring (SIM) neutral loss scanning corresponding to the loss Forebrain and hindbrain regions were identified50. Total CoA was determined of the phosphoadenosine diphosphate from CoA species. The ion source para- using 20–50 mg of tissue (liver, forebrain or hindbrain) homogenized in 2 ml of 1 meters were as follows: spray voltage 4000 V, capillary temperature 250 °C, capil- mM KOH. The pH was adjusted to 12.0 with 0.25 M KOH, and incubated at 55 °C lary offset −35 V, sheath gas pressure 10, auxiliary gas pressure 5, and tube lens for 2 h23. The pH of the sample was adjusted to 8.0, and the samples were deri- offset was set by infusion of the polytyrosine tuning and calibration in electrospray vatized with monobromobimane (mBBr) and analyzed by HPLC equipped with a mode. Acquisition parameters were as follows: scan time 0.5 s, collision energy 30 fluorescence detector. V, peak width Q1 and Q3 0.7 FWHM, Q2 CID gas 0.5 mTorr, source CID 10 V, neutral loss 507.0 m/z, SIM mass of 810 m/z with a scan width of 8 m/z to capture Generation of SynCre Pank1,Pank2 neuronal knockout mice.Generationofthe the signal from cellular acetyl-CoA and the [13C]acetyl-CoA (Sigma) internal fl fl fl fl Pank1 / and the Pank2 / mice was reported previously12,13.TheSynCre standard. transgene originated in B6.Cg-Tg(Syn1-cre)671Jxm/J transgenic mice (The Jackson Laboratory) that express the Cre recombinase driven by the synapsin1 PZ-2891 extraction and quantification by LC/MS/MS. Plasma (20 μl) was added (Syn) promoter. Pank1fl/fl,Pank2fl/+ SynCre+ (or Pank1fl/+,Pank2fl/fl SynCre+) to 100 μl acetonitrile containing 0.6 μM warfarin to a final concentration of 0.5 μM. females were mated with Pank1fl/fl,Pank2fl/fl SynCre− males to obtain PANK1fl/fl, The samples were incubated on ice for 30 min. Samples were spun at 3500×g for 10 PANK2fl/fl SynCre+ progeny that have both Pank1 and Pank2 conditionally

12 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE deleted in neuronal tissues. Control littermate mice had the Pank1fl/fl,Pank2fl/fl wave count (TWC) and the ELSD readings in LC/MS chromatogram (column: SynCre− genotype. PCR genotyping primer pairs and products are listed in Acquity BEH C18). 1H and 13C spectra were recorded using 400 or 500 MHz, using Supplementary Table 4. PCR analysis was used to genotype tail biopsies where a CDCl3 as solvent. The chemical shifts are reported in parts per million (ppm) fl 338 bp product indicated the oxed Pank1 allele, a 332 bp product indicated the relative to CDCl3. floxed Pank2 allele, and a 285 bp product indicated the presence of the Cre transgene. RNA was isolated from cryo-preserved liver or brain tissue. Synthesis of first-strand cDNA was obtained by reverse transcription using SuperScriptTM Synthesis of PZ-2891 and PZ-3067. 1-Boc-piperazine (200 mg, 1.43 mmol, 1 eq) RNase H reverse transcriptase, the RNA templates and random primers. and 6-chloropyridazine-3-carbonitrile (320 mg, 1.72 mmol, 1.2 eq) were dissolved Quantitatived real-time PCR was performed in triplicate using the ABI in anhydrous acetonitrile (6 ml) in microwave vial. Triethylamine (1 ml, 7.17 Prism®7700 Sequence Detection System with the primers listed in Supplemen- mmol, 5 eq) was added to it dropwise. The vial was capped and the reaction tary Table 5. The Taqman human GAPDH (Applied Biosystems) were used as mixture was heated at 160 °C for 30 min using microwave irradiation. The mixture 51 controls. All of the values were compared using the CT method ,andthe was cooled to room temperature and concentrated to dryness. The crude product amount of cDNA (2-ΔCT) was calculated relative to human glyceraldehyde-3- was dissolved in excess dichloromethane and washed with 1 M HCl followed by fi phosphate dehydrogenase mRNA. Neuron-speci c deletion of the Pank1-and water and brine. The organic phase was dried over MgSO4, and concentrated. The Pank2-floxed alleles was confirmed in brains by the presence of a 218 bp and purification was performed by flash column chromatography using hexane/ethyl 176 bp product, respectively, that were absent in liver (Supplementary Fig- acetate gradient to obtain tert-butyl 4-(6-cyanopyridazin-3-yl)piperazine-1-car- ure 14). Incomplete deletion of the floxed alleles in brain was indicated by boxylate (S1, 378 mg, 91% yield); 1H NMR (400 MHz, Chloroform-d) δ 7.47 (d, J coincidence of the Pank1-floxed and Pank1-deleted PCR products, and the = 9.6 Hz, 1 H), 6.84 (d, J = 9.6 Hz, 1 H), 3.80 (m, 4 H), 3.64–3.55 (m, 4 H), 1.49 (s, fl 13 δ Pank2- oxed and Pank2-deleted PCR products. Cell types other than neurons 9 H); C NMR (101 MHz, CDCl3) 158.65, 154.69, 130.85, 129.75, 116.81, 109.99, contributed to the undeleted, residual Pank1-andPank2-floxed genes in brain. 80.73, 44.39, 31.09, 28.53; LCMS (m/z): 290.16 (M+ + H, observed), 289.15 (M+, Wild-type matched control animals were derived from breeding pairs that were calculated). heterozygous for both Pank1-floxed and Pank2-floxed alleles and lacked the To deprotect the Boc functionality, S1 (105 mg, 0.363 mmol) was treated SynCre transgene. Following genotyping, SynCre+ and SynCre− Pank1-floxed with trifluoroacetic acid and dichloromethane (1:1, 3 ml) at room temperature. and Pank2-floxed mice regardless of sex were randomly enrolled into the After 1 h, reaction was monitored using UPLC and the mixture was treatment or control arms of the PZ-2891 trial as they emerged from the concentrated to dryness. The crude (6-(piperazin-1-yl)pyridazine-3- breeding program. carbonitrile)(S2)wasuseddirectlyfor the next step. 6-(piperazin-1-yl) pyridazine-3-carbonitrile (S2 crude, 0.363 mmol) was dissolved in 10 ml anhydrous dichloromethane. HATU (207 mg, 0.545 mmol, 1.5 eq) and DIPEA Open field locomotion . Individual mice were placed in an open rectangular arena (0.13ml,0.726mmol,2eq)wereaddedto it. The reaction mixture was stirred (36.8 cm × 43.2 cm) for 5 min during the light phase and motor activity was overnight at room temperature. The reaction mixture was then diluted with 1 M evaluated using a video tracking system provided by HVS Image with associated HCl solution and extracted with methylene chloride. The organic phase was 2100 Plus software (San Diego, CA, USA). Each subject was placed in the center of fi washed with water, brine, dried over MgSO4, ltered and concentrated. The the arena under standard overhead lighting and the total distance traveled and the crude residue was purified by flash column chromatography on a Biotage SP-1 percentage of time in motion were recorded. The spontaneous activity of each chromatography system using hexane/ethyl acetate to obtain 6-(4-(2-(4- subject in a novel open field provided a general measure of motor function of the + − fl isopropylphenyl)acetyl)piperazin-1-yl)pyridazine-3-carbonitrile (PZ-2891, 98 SynCre ( / ) Pank1, Pank2 oxed mice, age 45 days, treated or not with PZ-2891. mg, 77% yield for 2 steps); 1H NMR (400 MHz, Chloroform-d) δ 7.45 (d, J = 9.6 The behaviors of age-matched control animals, treated with PZ-2891 and Hz, 1 H), 7.19 (m, 4 H), 6.81 (d, J = 9.6Hz,1H),3.82(m,2H),3.76(s,2H), untreated, were quantified and compared. 3.75–3.66 (m, 4 H), 3.63 (m, 2 H), 2.88 (m, 1 H), 1.23 (d, J = 6.9Hz,6H); 13C δ NMR (126 MHz, CDCl3) 170.27, 158.54, 147.93, 131.79, 130.91, 130.02, Crystallization and structure determination. PANK3 with two amino acids 128.59, 127.16, 116.68, 110.05, 45.42, 44.54, 44.01, 41.12, 40.85, 33.87, 24.11; + + (DD) added to the carboxy-terminus was expressed, purified and crystalized9. HRMS (m/z): 350.1983 (M + H, observed), 349.1903 (M ,calculated).1H- The crystals of the PANK3•AMPPNP•Mg2+•pantothenate complex9 were NMR and 13C-NMR spectra of PZ-2891 are in Supplementary Fig. 20. Liquid soaked in mother liquor (0.2 M ammonium acetate, 0.1 M citrate, pH, 5.6, 50 chromatography analysis and mass spectrum of PZ-2891 are in Supplementary mM MgCl2, 32% polyethylene glycol (4 K), 4% DMSO, 10 mM AMPPNP Fig. 21. (adenosine 5′-(β,γ-imido)triphosphate) and 1 mM PZ-2891 for two days. The 6-(Piperazin-1-yl)pyridazine-3-carbonitrile (S2 crude, 0.287 mmol) was crystals were cryoprotected with 29% ethylene glycol. Diffraction data were dissolved in anhydrous dichloromethane (10 ml) which was followed by addition of collected at the SER-CAT beam line 22-ID at the Advanced Photon Source, and 2-bromo-1-(4-isopropylphenyl)ethan-1-one (104 mg, 0.430 mmol, 1.5 eq) and processed using HKL200052. The structure was solved by molecular replacement DIPEA (0.25 ml, 1.434 mmol, 5 eq). The reaction mixture was stirred at room using the PANK3 structure (PDB ID: 3SMS) and the program PHASER53.The temperature for 3 h. After the reaction was completed, 1 M HCl was added and it 54 55 structure was refinedandoptimizedusingPHENIX and COOT , respectively. was extracted with dichloromethane. The organic phase was dried with MgSO4, The refined structure was validated using MolProbity56. The atomic coordinates and the solvent was removed. The purification was performed by flash column and structure factors have been deposited in the Protein Data Bank as PDB entry chromatography using hexane/ethyl acetate gradient to obtain product 6-(4-(2-(4- 6B3V. The data collection and refinement statistics are presented in Supple- isopropylphenyl)-2-oxoethyl)piperazin-1-yl)pyridazine-3-carbonitrile (PZ-3067, mentary Table 3. Ramachandran statistics (favored/allowed/outliers (%)) 98.28/ 92 mg, 92% yield). 1H NMR (400 MHz, Chloroform-d) δ 7.93 (d, J = 8.3 Hz, 2 H), 1.72/0.0 for the PZ-2891 complex. All structures were rendered with PyMOL 7.44 (d, J = 9.6 Hz, 1 H), 7.33 (d, J = 8.3 Hz, 2 H), 6.83 (d, J = 9.5 Hz, 1 H), 3.89 (m, (version 1.8, Schrödinger, LLC). For co-crystallization of the PAN- 6 H), 2.98 (h, J = 6.9 Hz, 1 H), 2.77 (m, 4 H), 1.27 (d, J = 6.9 Hz, 6 H); 13C NMR 2+ δ K3•AMPPNP•Mg •PZ-2891 complex, the protein solution was incubated with (101 MHz, CDCl3) 195.49, 158.63, 155.37, 133.78, 130.73, 129.42, 128.46, 126.95, ′ β γ 116.97, 109.83, 63.94, 52.94, 44.55, 34.47, 23.79; HRMS (m/z): 350.1985 (M+ + H, 25 mM AMPPNP (adenosine 5 -( , -imido)triphosphate), 25 mM MgCl2,and + 1.25 mM PZ-2891 (final 5% DMSO) to form a complex. The ternary complex observed), 349.1903 (M , calculated). was crystallized using the hanging-drop vapor-diffusion method at 18 °C. Crystalsweregrownbymixing1.5μl of complex solution with 1.0 μl of reservoir solution containing 0.1 M HEPES, pH 7.5, 10% isopropanol, 20% PEG 4 K, and Synthesis of PZ-2789. 6-(Piperazin-1-yl)nicotinonitrile (102 mg, 0.542 mmol, 1 0.1 M guanidine hydrochloride. eq) was dissolved in dichloromethane (5 ml) and 1-(tert-butyl)-4-iso- cyanatobenzene (104 mg, 0.596 mmol, 1.1 eq) was added to it which was followed by dropwise addition of triethylamine (0.08 ml, 0.596 mmol, 1.1 eq). The reaction Statistics. All statistical calculations were performed using GraphPad Prism. Tests mixture was stirred at room temperature and the reaction was monitored using ’ between two groups used two-tailed unpaired Student s t test. Data are presented as LC-MS. After the reaction was completed (within 1 h), the reaction mixture was means ± SD, and the p values and number of biological replicates are provided in evaporated to dryness. The crude residue was purified by flash column chroma- fi the gures. The log-rank (Cox-Mantel) test was used to analyze the survival data. tography using hexane/ethyl acetate gradient to obtain N-(4-(tert-butyl)phenyl)-4- The pre-established exclusion criteria in the Grubbs test was used to determine if (5-cyanopyridin-2-yl)piperazine-1-carboxamide (PZ-2789, 165 mg, 84% yield); 1H there were outliers. Sample size (at least 5 mice per group) was chosen based on NMR (400 MHz, Chloroform-d) δ 8.46 (dd, J = 2.4, 0.8 Hz, 1 H), 7.69 (dd, J = 9.0, pilot experiments measuring the normal distribution of CoA levels in mouse tissues 2.3 Hz, 1 H), 7.38–7.32 (m, 2 H), 7.30 (d, J = 2.1 Hz, 2 H), 6.63 (dd, J = 9.0, 0.9 Hz, (between 4% and 10%). No mice were excluded from the study, and there was no 1 H), 6.31 (s, 1 H), 3.83 (m, 4 H), 3.74–3.65 (m, 4 H), 1.32 (s, 9 H); 13C NMR (101 blinding. δ MHz, CDCl3) 155.21, 152.77, 146.72, 140.23, 135.95, 125.99, 120.16, 105.84, 97.27, 43.93, 43.28, 34.45, 31.54; HRMS (m/z): 364.2152 (M+ + H, observed), + General chemical methods. Unless otherwise noted, all reactions were carried out 363.2059 (M , calculated). in flame-dried glassware under a static nitrogen atmosphere with anhydrous sol- vent. All reagents were obtained from commercially available sources and used without purification. Purification was handled by flash Silica Gel chromatography Data availability using a Biotage SP-1 chromatography system and SNAP KP-Sil cartridges. All The PANK3•AMPPNP•Mg2+•PZ-2891 crystal structure and diffraction data have been final compounds were purified to >95% purity indicated as averages of the total deposited with the worldwide protein data bank under accession code 6B3V [https://doi.

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 org/10.2210/pdb6B3V/pdb]. The data that support the findings of this study are available 25. Di Meo, I. et al. Acetyl-4’-phosphopantetheine is stable in serum and prevents from the corresponding authors on request. phenotypes induced by pantothenate kinase deficiency. Sci. Rep. 7, 11260 (2017). Received: 13 March 2018 Accepted: 20 September 2018 26. Brunetti, D. et al. Pantethine treatment is effective in recovering the disease phenotype induced by ketogenic diet in a pantothenate kinase-associated neurodegeneration mouse model. Brain 137,57–68 (2014). 27. Rana, A. et al. Pantethine rescues a Drosophila model for pantothenate kinase- associated neurodegeneration. Proc. Natl Acad. Sci. USA 107, 6988–6993 (2010). 28. Orellana, D. I. et al. Coenzyme A corrects pathological defects in human References neurons of PANK2-associated neurodegeneration. EMBO Mol. Med. 8, 1. Zhou, B. et al. A novel pantothenate kinase gene (PANK2) is defective in 1197–1211 (2016). – Hallervorden-Spatz syndrome. Nat. Genet. 28, 345 349 (2001). 29. Zizioli, D. et al. Knock-down of pantothenate kinase 2 severely affects the 2. Schneider, S. A. et al. Genetics and pathophysiology of neurodegeneration development of the nervous and vascular system in zebrafish, providing new – with brain iron accumulation (NBIA). Curr. Neuropharmacol. 11,59 79 insights into PKAN disease. Neurobiol. Dis. 85,35–48 (2016). (2013). 30. Khatri, D. et al. Down-regulation of coasy, the gene associated with NBIA-VI, fl 3. Hay ick, S. J. et al. Genetic, clinical, and radiographic delineation of reduces Bmp signaling, perturbs dorso-ventral patterning and alters neuronal – Hallervorden-Spatz syndrome. N. Engl. J. Med. 348,33 40 (2003). development in zebrafish. Sci. Rep. 6, 37660 (2016). 4. Hogarth, P. et al. Consensus clinical management guideline for pantothenate 31. Hitchcock, S. A. & Pennington, L. D. Structure–brain exposure relationships. kinase-associated neurodegeneration (PKAN). Mol. Genet. Metab. 120, J. Med. Chem. 49, 7559–7583 (2006). – 278 287 (2017). 32. Rankovic, Z. CNS drug design: balancing physicochemical properties for 5. Leonardi, R., Zhang, Y.-M., Rock, C. O. & Jackowski, S. Coenzyme A: back in optimal brain exposure. J. Med. Chem. 58, 2584–2608 (2015). – action. Prog. Lipid Res. 44, 125 153 (2005). 33. Leonardi, R. et al. Modulation of pantothenate kinase 3 activity by small 6. Rock, C. O., Calder, R. B., Karim, M. A. & Jackowski, S. Pantothenate kinase molecules that interact with the substrate/allosteric regulatory domain. Chem. regulation of the intracellular concentration of coenzyme A. J. Biol. Chem. Biol. Drug. Des. 17, 892–902 (2010). – 275, 1377 1383 (2000). 34. Sharma, L. K. et al. A high-throughput screen reveals new small-molecule 7. Zhang, Y.-M., Rock, C. O. & Jackowski, S. Feedback regulation of murine activators and inhibitors of pantothenate kinases. J. Med. Chem. 58, pantothenate kinase 3 by coenzyme A and coenzyme A thioesters. J. Biol. 1563–1568 (2015). – Chem. 280, 32594 32601 (2005). 35. Hopkins, A. L., Keseru, G. M., Leeson, P. D., Rees, D. C. & Reynolds, C. H. 8. Hong, B. S. et al. Crystal structures of human pantothenate kinases. Insights The role of ligand efficiency metrics in drug discovery. Nat. Rev. Drug. Discov. into allosteric regulation and mutations linked to a neurodegeneration 13, 105–121 (2014). – disorder. J. Biol. Chem. 282, 27984 27993 (2007). 36. Hann, M. M. & Keseru, G. M. Finding the sweet spot: the role of nature and 9. Subramanian, C. et al. Allosteric regulation of mammalian pantothenate nurture in medicinal chemistry. Nat. Rev. Drug. Discov. 11, 355–365 (2012). – kinase. J. Biol. Chem. 291, 22302 22314 (2016). 37. Copeland, R. A. Drug-target resisdence time. in Evaluation of Enzyme 10. Dansie, L. E. et al. Physiological roles of the pantothenate kinases. Biochem. Inhibitors in Drug Discovery: A Guide for Medicinal Chemists and – Soc. Trans. 42, 1033 1036 (2014). Pharmacologists 2nd edn 287-344 (John Wiley and Sons, Hoboken, New 11. Zhang, Y. M. et al. Chemical knockout of pantothenate kinase reveals the Jersey, 2013). metabolic and genetic program responsible for hepatic coenzyme A 38. Jafari, R. et al. The cellular thermal shift assay for evaluating drug target – homeostasis. Chem. Biol. 14, 291 302 (2007). interactions in cells. Nat. Protoc. 9, 2100–2122 (2014). 12. Leonardi, R., Rehg, J. E., Rock, C. O. & Jackowski, S. Pantothenate kinase 1 is 39. Corbin, D. R. et al. Excess coenzyme A reduces skeletal muscle performance required to support the metabolic transition from the fed to the fasted state. and strength in mice overexpressing human PANK2. Mol. Genet. Metab. 120, PLoS ONE 5, e11107 (2010). 350–362 (2017). 13. Garcia, M., Leonardi, R., Zhang, Y. M., Rehg, J. E. & Jackowski, S. Germline 40. Bettencourt, C. et al. Gene co-expression networks shed light into diseases of deletion of pantothenate kinases 1 and 2 reveals the key roles for CoA in brain iron accumulation. Neurobiol. Dis. 87,59–68 (2016). postnatal metabolism. PLoS ONE 7, e40871 (2012). 41. Cowan-Jacob, S. W., Jahnke, W. & Knapp, S. Novel approaches for targeting fi 14. Leonardi, R., Rock, C. O. & Jackowski, S. Pank1 deletion in leptin-de cient kinases: allosteric inhibition, allosteric activation and pseudokinases. Future fi mice reduces hyperglycaemia and hyperinsulinaemia and modi es global Med. Chem. 6, 541–561 (2014). – metabolism without affecting insulin resistance. Diabetologia 57, 1466 1475 42. Rettenmaier, T. J. et al. A small-molecule mimic of a peptide docking motif (2014). inhibits the protein kinase PDK1. Proc. Natl Acad. Sci. USA 111, 18590–18595 15. Dusi, S. et al. Exome sequence reveals mutations in CoA synthase as a cause of (2014). neurodegeneration with brain iron accumulation. Am. J. Hum. Genet. 94, 43. Breus, O., Panasyuk, G., Gout, I. T., Filonenko, V. & Nemazanyy, I. CoA – 11 22 (2014). synthase is phosphorylated on tyrosines in mammalian cells, interacts with 16. Zhang, Y.-M., Rock, C. O. & Jackowski, S. Biochemical properties of human and is dephosphorylated by Shp2PTP. Mol. Cell Biochem. 335, 195–202 pantothenate kinase 2 isoforms and mutations linked to pantothenate kinase- (2010). – associated neurodegeneration. J. Biol. Chem. 281, 107 114 (2006). 44. Daugherty, M. et al. Complete reconstitution of the human coenzyme A fi 17. Hörtnagel, K., Prokisch, H. & Meitinger, T. An isoform of hPANK2, de cient biosynthetic pathway via comparative genomics. J. Biol. Chem. 277, in pantothenate kinase-associated neurodegeneration, localizes to 21431–21439 (2002). – mitochondria. Hum. Mol. Genet. 12, 321 327 (2003). 45. Ramaswamy, G., Karim, M. A., Murti, K. G. & Jackowski, S. PPARa controls 18. Alfonso-Pecchio, A., Garcia, M., Leonardi, R. & Jackowski, S. the intracellular coenzyme A concentration via regulation of PANK1a gene Compartmentalization of mammalian pantothenate kinases. PLoS ONE 7, expression. J. Lipid Res. 45,17–31 (2004). e49509 (2012). 46. Shumar, S. A. et al. Nudt19 is a renal CoA diphosphohydrolase with 19. Johnson, M. A. et al. Mitochondrial localization of human PANK2 and biochemical and regulatory properties that are distinct from the hepatic Nudt7 hypotheses of secondary iron accumulation in pantothenate kinase-associated isoform. J. Biol. Chem. 293, 4134–4148 (2018). – neurodegeneration. Ann. N. Y. Acad. Sci. 1012, 282 298 47. Shumar, S. A. et al. Induction of neuron-specific degradation of Coenzyme A (2004). models pantothenate kinase-associated neurodegeneration by reducing motor 20. Brunetti, D. et al. Pantothenate kinase-associated neurodegeneration: altered coordination in mice. PLoS ONE 10, e0130013 (2015). mitochondria membrane potential and defective respiration in Pank2 knock- 48. Morrison, J. F. Kinetics of the reversible inhibition of enzyme-catalysed – out mouse model. Hum. Mol. Genet. 21, 5294 5305 (2012). reactions by tight-binding inhibitors. Biochim. Biophys. Acta 185, 269–286 21. Leonardi, R., Zhang, Y. M., Lykidis, A., Rock, C. O. & Jackowski, S. (1969). Localization and regulation of mouse pantothenate kinase 2. FEBS Lett. 581, 49. Stella, V. J. & He, Q. Cyclodextrins. Toxicol. Pathol. 36,30–42 (2008). – 4639 4644 (2007). 50. Paxinos, G. & Franklin, K. B. J. The Mouse Brain in Stereotaxic Coordinates fi 22. Kuo, Y. M. et al. De ciency of pantothenate kinase 2 (Pank2) in mice leads to (Elsevier Science, San Diego, USA, 2004). – retinal degeneration and azoospermia. Hum. Mol. Genet. 14,49 57 (2005). 51. Winer, J., Jung, C. K., Shackel, I. & Williams, P. M. Development and 23. Zano, S. P., Pate, C., Frank, M., Rock, C. O. & Jackowski, S. Correction of a validation of real-time quantitative reverse transcriptase-polymerase chain fi genetic de ciency in pantothenate kinase 1 using phosphopantothenate reaction for monitoring gene expression in cardiac myocytes in vitro. Anal. – replacement therapy. Mol. Genet. Metab. 116, 281 288 (2015). Biochem. 270,41–49 (1999). ’ 24. Srinivasan, B. et al. Extracellular 4 -phosphopantetheine is a source for 52. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in – intracellular coenzyme A synthesis. Nat. Chem. Biol. 11, 784 792 (2015). oscillation mode. Methods Enzymol. 276, 307–326 (1997).

14 NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06703-2 ARTICLE

53. McCoy, A. J. Solving structures of protein complexes by molecular Additional information replacement with Phaser. Acta Crystallogr. D. Biol. Crystallogr. 63,32–41 Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- (2007). 018-06703-2. 54. Afonine, P. V. et al. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr. D. Biol. Crystallogr. 68, 352–367 (2012). Competing interests: The authors declare the following competing interests. L.K.S., C. 55. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development S., M.-K.Y., C.O.R., R.E.L. and S.J. are inventors on a pending patent application of Coot. Acta Crystallogr. D. Biol. Crystallogr. 66, 486–501 (2010). (PCT/US17/39037) “Small molecule modulators of pantothenate kinases” held by St. 56. Chen, V. B. et al. MolProbity: all-atom structure validation for Jude Children’s Research Hospital that covers the pantazine chemical series disclosed macromolecular crystallography. Acta Crystallogr. D. Biol. Crystallogr. 66, in this manuscript. S.J., R.E.L., C.O.R. and S.W.W. received research funding from 12–21 (2010). CoA Ttherapeutics. The authors declare no other competing interests.

Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions/ Acknowledgements This work was supported by a sponsored research grant from CoA Therapeutics, Cancer Publisher's note: Springer Nature remains neutral with regard to jurisdictional Center core grant CA21765 and the American Syrian Lebanese Associated Charities. We claims in published maps and institutional affiliations. thank the St. Jude Protein Production Facility for protein expression/purification, the St. Jude Hartwell Center for DNA sequencing, Louis Richmond for animal experiments, Caroline Pate for CETSA experiments, Katie Creed for biochemical and biophysical fi assays, Karen Miller for animal experiments, protein puri cation and molecular biology, Open Access This article is licensed under a Creative Commons Jina Wang for cell culture experiments, Brett Waddell of the Molecular Interaction Attribution 4.0 International License, which permits use, sharing, Shared Resource for surface plasmon resonance analysis, and Lei Yang of the High adaptation, distribution and reproduction in any medium or format, as long as you give Throughput Analytical Core for ADME analysis. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Author contributions indicated otherwise in a credit line to the material. If material is not included in the All authors contributed to formulating the conclusions, writing the manuscript and article’s Creative Commons license and your intended use is not permitted by statutory approved the text, tables and figures. L.K.S. and R.E.L. designed and interpreted the regulation or exceeds the permitted use, you will need to obtain permission directly from chemical biology experiments. L.K.S. synthesized the compounds and prepared the the copyright holder. To view a copy of this license, visit http://creativecommons.org/ formulation for animal studies. S.W.W. and M-K.Y. designed and interpreted the licenses/by/4.0/. structural biology experiments. M.W.F. performed the CoA measurements and mass spectrometry. C.O.R. and C.S. designed and interpreted the biochemistry experiments, S. J. derived the animal model and S.J. and C.S. designed and interpreted the cell and © The Author(s) 2018 animal experiments.

NATURE COMMUNICATIONS | (2018) 9:4399 | DOI: 10.1038/s41467-018-06703-2 | www.nature.com/naturecommunications 15